BiCAT builds bimanual coordination directly into VLA action tokens

DJiafei · x · 2026-11-19

DJiafei introduces BiCAT (Bimanual Coordination Action Tokenization), a method that builds coordination directly into a VLA's action tokens.

The core argument: coordinating two arms takes more than knowing what to do — coordination must be modeled at the action-representation level. BiCAT suits high-demanding tasks requiring coordination and low latency (e.g. two-arm ball-handling demos).

Related event: BiCAT Encodes Bimanual Coordination Directly into VLA Action Tokens(10 posts)→

Original post →

More from Embodied

Embodied channel →